Rendering-Aware Bayesian 3D Gaussian Splatting with Native Uncertainty and Adaptive Complexity Control
Quick Answer
The proposed Bayesian 3D Gaussian Splatting (3DGS) framework enhances real-time novel-view synthesis by incorporating uncertainty and adaptive complexity control, achieving a PSNR improvement of +0.453 dB and a 17x reduction in coverage error compared to traditional methods.
Quick Take
The proposed Bayesian 3D Gaussian Splatting (3DGS) framework enhances real-time novel-view synthesis by incorporating uncertainty and adaptive complexity control, achieving a PSNR improvement of +0.453 dB and a 17x reduction in coverage error compared to traditional methods. This positions Bayesian 3DGS as a viable solution for active view selection tasks, outperforming existing models with lower training costs.
Key Points
- Bayesian 3DGS uses Normal-Inverse-Wishart posterior for Gaussian geometry tracking.
- Achieves +0.453 dB PSNR improvement in active-view tasks with fixed budgets.
- Reduces 95% coverage error by about 17x compared to shared proxies.
- Outperforms PPU-style and NIW-proxy acquisition methods.
- Requires roughly one-third the training cost of traditional methods.
Paper Resources
📖 Reader Mode
~2 min readAbstract:3D Gaussian splatting (3DGS) is a strong representation for real-time novel-view synthesis, but its standard training pipeline relies on point estimates and hand-tuned heuristics, providing no native uncertainty or principled complexity control. This is most limiting under sparse views or fixed acquisition budgets, where a model must identify weakly supported geometry and select informative views. We introduce a rendering-aware Bayesian 3DGS framework that tracks Gaussian geometry with a Normal-Inverse-Wishart posterior over means and covariances using renderer-derived surrogate summaries. An optional Dirichlet-process extension adds a probabilistic component-usage signal, and the training schedule makes the closed-form versus approximate inference boundary explicit. Re-rendering posterior geometry samples yields native predictive uncertainty for interval calibration and active view selection. In a fixed-budget 16-to-32 active-view task, native NIW acquisition improves PSNR by +0.453 dB and LPIPS by -0.0146 over a scoring-only 3-member standard-ensemble baseline, winning 29/39 scene-seed pairs and 10/13 scene means; it also improves over PPU-style (+0.355 dB) and NIW-proxy (+0.401 dB) acquisition. NIW native intervals reduce 95% coverage error by about 17x relative to a shared proxy (0.046 vs. 0.796) and are about 10x closer to nominal coverage than a 3-member deep ensemble (0.047 vs. 0.454) at roughly one-third the training cost. As a reconstruction compatibility check, paired NIW-vs-standard analysis over 39 scene-seed runs yields +0.030 dB PSNR with 1.6% additional training time. These results position Bayesian 3DGS as a practical probabilistic scene representation for decision-facing tasks such as active view selection.
| Comments: | 26 pages, 4 figures, 24 tables including appendix. Preprint |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2607.05522 [cs.CV] |
| (or arXiv:2607.05522v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2607.05522 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Gaoxiang Jia [view email]
[v1]
Mon, 6 Jul 2026 18:01:08 UTC (178 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →LLM-Guided ANN Index Optimization for Human-Object Interaction Retrieval
A phase-aware LLM agent optimizes human-object interaction retrieval, outperforming Optuna TPE by 33.3% and VDTuner by 34.2% on the HICO-DET benchmark. This method enhances throughput by 15.3x over UniIR and demonstrates strong transferability across vector database management systems.